AI is on the rise. So are AI safety and ethical AI.

Artificial Intelligence - AI in short - is a broad discipline whose goal is to create intelligent machines by emulating the full range of human cognition. Given its ambition to build machines that potentially exceed the full range of human abstraction, AI will have an ever-growing impact on our society. As a result, ethical AI systems and AI safety are bound to become hot topics.

The Rise of the Machines

For sure, AI is no science fiction but already has a significant impact on our lives today. Just think of using your face to access your smartphone or your voice to instruct a virtual assistant. Not to mention the numerous recommendations you get for food, clothing, travel destinations, what to watch or listen to, based on how you visit a website or use an app. The State of AI Report 2021 discusses the most recent advances in AI and outlines how AI is stepping up in more concrete ways throughout our society.

All this is powered by the very data we generate ourselves, combined with some form of Machine Learning (a subset of AI that gives machines the ability to learn from data without being explicitly programmed to do so).

Is AI coming for me?

As Artificial Intelligence also advances in the workplace, various ominous predictions about its potential dangers (“AI is coming for your job”) start to circulate in the media – even though these are usually fueled by fear and fantasy rather than fact.

Surely, we shouldn’t turn a blind eye to worries about the long-term implications of AI developments. Nor should we ignore questions on AI safety and ethical AI systems. In the first place, however, we need to recognize how AI helps reduce the drudgery involved in tedious routine tasks and enables things that weren't possible so far. Just think of some recent AI-powered applications with a very positive impact on our society:

• A system that can detect employee health and safety incidents in real-time is now live in over 30 cities.
• Insurance companies use AI-augmented systems to assist homeowners in predicting repair costs and unlocking faster insurance claim pay-outs for damages done to their home after a natural disaster.
• The UK national grid uses a transformer-based system to forecast electricity supply and demand, which is essential to running the grid on net-zero generation by 2025.
• Google Maps received over 100 new AI features such as indoor navigation with Augmented Reality and a new routing option that optimizes for lower fuel consumption and CO₂ emissions.
• Online grocers can now use AI models to automate up to 98% of their stock replenishment decisions to increase product availability and reduce waste.

How to design broadly adopted ethical AI systems?

Despite the increasing number of AI breakthroughs that better our lives, it is fair to claim that the development of new AI applications has primarily been driven by geopolitical and economic competition. Moral and ethical concerns have had minimal impact on AI developments so far. But as companies and governments are increasingly investing in AI developments, experts and advocates worldwide become concerned about the long-term effects of AI applications.

Those experts and advocates are mainly worried about how advances in Artificial Intelligence will affect what it means to be human, and their surging worries do deserve careful attention. One of the main challenges in addressing those worries is to design broadly adopted ethical AI systems. That is challenging because ethics are hard to define and because social standards evolve as cultures change. Moreover, since lots of historical data are needed to train Artificial Intelligence algorithms on ethical behavior, the available data may be largely insufficient to represent our current ethical standards.

AI safety and ethical AI are NOT a story of gloom and doom

On the bright side of things, a survey conducted by the Pew Research Center among researchers who published in top Machine Learning conferences showed that 68% of them are convinced that AI safety needs to be more prioritized than it is today. AI safety stands for the complete set of undertakings to ensure that Artificial Intelligence is deployed in ways that do not harm humankind.

It is one of the many signals confirming the increasing importance attached to AI safety and ethical AI systems. As new initiatives and organizations with an explicit focus on those topics continue to spring up (led by talented researchers), one thing is for sure: AI safety and ethical AI are NOT a story of gloom and doom.